Abstract: An improved method for interconnect delay analysis for VLSI circuits reduces a parasitic graph for moment computation by eliminating one or more nodes in the graph. The elimination process is performed based upon the degree of the nodes. By eliminating nodes in this fashion, the computation complexity is significantly reduced. With this elimination process, resistor loops and crossed loops can also be solved. The order in which the nodes are eliminated is optimized using the depth-first-search method on the parasitic graphs, further reducing the computation complexity. The method provides a consistent functional interface, applicable to different circuit model structures. In addition, the method accounts for coupling capacitance between interconnects.

Abstract: A method of calculating delay for a process variation includes finding a value for each of exactly two independent variables that results in a maximum or minimum variation of estimated cell delay plus net delay, calculating a variation of resistance from the value found for each of the exactly two independent variables, calculating a variation of capacitance from the value found for each of the exactly two independent variables, adding the calculated variation of resistance to a net resistance to generate a modified net resistance for a selected net, adding the calculated variation of capacitance to a net capacitance to generate a modified net capacitance for the selected net, and calculating the cell delay plus net delay from the modified net resistance and the modified net capacitance.

Abstract: A method of modeling delays in an integrated circuit design is disclosed that may be used to reduce the computation time of path delays in an integrated circuit design. A method of modeling delays in an integrated circuit design includes the steps of receiving as input a description of an integrated circuit design; identifying at least one small net in the integrated circuit design from the description; approximating an effective capacitance of the at least one small net by the total capacitance; and approximating an interconnect delay of the at least one small net by zero.